path: root/itches/emacs/tracker.org

Since my contributions to "GNU Emacs" are spread over multiple
repositories and bug tracking systems, I figured I would aggregate
them in my own "tracker".  This allows me to keep track of what I'm
working on; it can also double as a [[https://jvns.ca/blog/brag-documents/][brag document]] when filtered on
=PatchApplied= properties.

Speaking of properties, here are the ones I'm attaching to these
entries:

- =Role= ::
  - =author= :: for issues I reported,
  - =watcher= :: issues reported by someone else…
    - I have taken a stab at,
    - I could take a stab at, if time permits,
    - I'll tip my hat to whoever solves it.
- =PatchApplied= :: whether my patches were applied,
- =TestAdded= :: for bug reports, whether I added a unit test.

* GNU Emacs
** Build process
*** Dump
**** DONE [[bug:38492]] Warn pdumper users when pure space has been overflowed
:PROPERTIES:
:Role:           author
:END:
Fixed by Eli.
*** LEIM
**** DONE [[bug:54816]] Silent "Compiler-macro error for cl-member" when building from scratch
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
** Core
*** Display
**** Cairo support
***** DONE [[bug:35781]] Improve font display on Cairo builds
:PROPERTIES:
:Role:           author
:END:
Fixed by Yamamoto Mitsuharu.
***** DONE [[bug:35871]] Fix SVG transparency with Cairo
:PROPERTIES:
:Role:           author
:END:
Fixed by Yamamoto Mitsuharu.
**** Fonts
***** TODO Rank candidate fonts consistently wrt. their version field
I sometimes have two versions of a font installed, e.g. the
distro-provided one, plus an upstream one tucked in =~/.local=.

Emacs seems to not use the same heuristics as fontconfig's ~fc-match~,
and sometimes picks the older version.

No bug filed yet; discussed first on [[https://lists.gnu.org/archive/html/help-gnu-emacs/2022-01/msg00042.html][help-gnu-emacs]] and [[https://lists.gnu.org/archive/html/emacs-devel/2022-01/msg00725.html][emacs-devel]],
then brought up on the [[https://lists.freedesktop.org/archives/fontconfig/2022-January/006859.html][fontconfig mailing list]].  Tentative conclusion:
fontconfig's ~fc-match~ seems to take =FC_FONTVERSION= into account
when ranking fonts; it'd be nice if Emacs's ~find-font~ did so too.
**** "Pure GTK" port
***** DONE [[https://lists.gnu.org/archive/html/emacs-devel/2021-12/msg01845.html][emacs-devel:<87wnk0wyi7.fsf@gmail.com>]] Let =--with-webp= work with =--with-pgtk=
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
**** DONE [[bug:41584]] org-indent-mode's line-prefix text property flickers near overlays
:PROPERTIES:
:Role:           author
:END:
First reported as [[https://github.com/dgutov/diff-hl/issues/142][dgutov/diff-hl#142]].  Fixed by Eli.
**** TODO [[bug:52983]] Implement visual-fill-column equivalent in display engine
The status quo is disheartening:

- hard-wrapping is illegible in narrow windows,
- soft-wrapping is illegible in wide windows.

This feature would allow unwrapped text to have a reasonable upper
limit on line length, with smart wrapping prefixes using
=adaptive-wrap=; even hard-wrapped text would not look /too/ ugly in
narrow windows.
**** TODO Line height changes depending on visible characters
E.g. with wide Org tables and ~truncate-lines~ enabled, if the
~org-table~ face is set to a smaller height, then scrolling
horizontally makes the lines "shake" because the final newline does
not have the ~org-table~ face, and is thus taller.

Not sure what the right approach is:
1. ask Org to fontify the final newline,
2. ask display engine to consider the whole line to compute the height
3. ask display engine to ignore the newline to compute the height
*** Input
**** TODO Translate unshifted keys to shifted if no bindings are found
To make =C-x [0-9]= more accessible on AZERTY.  Firefox does this, cf
[[https://hg.mozilla.org/mozilla-unified/file/FIREFOX_80_0_1_RELEASE/widget/gtk/nsGtkKeyUtils.cpp#l1207][here]]:

#+begin_src cpp
  // Retry with shifted keycode.
  guint shiftState = (baseState | keymapWrapper->GetModifierMask(SHIFT));
  uint32_t shiftedChar = keymapWrapper->GetCharCodeFor(aGdkKeyEvent, shiftState,
                                                       aGdkKeyEvent->group);
  if (IsBasicLatinLetterOrNumeral(shiftedChar)) {
    // A shifted character can be an ASCII alphabet on Hebrew keyboard
    // layout. And also shifted character can be an ASCII numeric on
    // AZERTY keyboad layout.  Then, it's a good hint for deciding our
    // keyCode.
    return WidgetUtils::ComputeKeyCodeFromChar(shiftedChar);
  }
#+end_src

and [[https://hg.mozilla.org/mozilla-unified/file/FIREFOX_80_0_1_RELEASE/widget/gtk/nsGtkKeyUtils.cpp#l1896][there]]:

#+begin_src cpp
uint32_t KeymapWrapper::GetCharCodeFor(const GdkEventKey* aGdkKeyEvent,
                                       guint aModifierState, gint aGroup) {
  guint keyval;
  if (!gdk_keymap_translate_keyboard_state(
          mGdkKeymap, aGdkKeyEvent->hardware_keycode,
          GdkModifierType(aModifierState), aGroup, &keyval, nullptr, nullptr,
          nullptr)) {
    return 0;
  }
  GdkEventKey tmpEvent = *aGdkKeyEvent;
  tmpEvent.state = aModifierState;
  tmpEvent.keyval = keyval;
  tmpEvent.group = aGroup;
  return GetCharCodeFor(&tmpEvent);
}
#+end_src

Maybe look at ~lookup-key~ in ~src/keymap.c~?  Although
~src/gtkutil.c~ seems to be the place making the most calls to ~gdk_~
functions.
*** Native compilation
**** DONE [[bug:41077]] Segfaults when compiling ELC+ELN
:PROPERTIES:
:Role:           author
:END:
Took a couple of tries (and 3 days straight of compilation), but I got
the branch to compile on my Samsung NC10.  See also update 8 on
Andrea's [[https://akrl.sdf.org/gccemacs.html][progress page]].
**** DONE [[bug:41194]] fibn benchmark exhausts memory
:PROPERTIES:
:Role:           author
:END:
Fixed by Andrea (see update 9 on his [[https://akrl.sdf.org/gccemacs.html][progress page]]).
** Elisp
*** DONE [[bug:30008]] Subdirectory vs major mode in .dir-locals.el
:PROPERTIES:
:Role:           author
:END:
Unnoticed for a year, then merged with subsequent duplicate [[bug:33400]].
*** TODO [[bug:31586]] ~frame-title-format~ doesn't save match data
:PROPERTIES:
:Role:           watcher
:END:
*** DONE [[bug:33697]] file-truename messes with match data
:PROPERTIES:
:Role:           author
:END:
Merged with [[bug:31586]], a more general issue dealing with match data
being "poisoned" by a user's =:eval= form in frame-title-format.
*** DONE [[bug:36803]] Update mode-line of every window when compilation ends
:PROPERTIES:
:Role:           author
:END:
Fixed by Stefan.
*** DONE [[bug:57733]] Fix ~replace-*-in-region~ crash
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
*** TODO Add command to toggle password visibility
**** cover letter
Hello Emacs,

On occasion, e.g. when saving .authinfo.gpg, I'd like to briefly peek at
the password I'm typing.  Entering the password twice is an appreciated
safeguard, but I know I'm perfectly able to make the same typo twice in
a row; I'd love for read-passwd to offer a command to toggle the
password visibility.

(I'm dismissing the "write it in the clear, then cut it and paste it"
solution, since I don't want to add the kill ring nor my desktop
environment's clipboard manager to the equation)

I'm attaching a proof-of-concept.  Additional considerations, in no
particular order:

- For the binding, I wanted something easy to mash quickly, but
  something more mnemonic might be in order?

- Some applications out there show the password while the keyboard
  shortcut (or the mouse button) remains pressed, and hide it back as
  soon as it is released; I like the idea, but I have no idea whether
  that's feasible with Emacs.

- The (member …) check could be replaced with e.g. (get-char-property
  (minibuffer-prompt-end) 'display); that's how I implemented it at
  first.  It would fail to turn hiding off when the user has not entered
  anything though.

- I don't know how much documentation this new command would warrant:

    - Should it be mentioned in read-passwd's docstring?  C-u isn't now,
      but it was way back when, before 2012-04-11 "* lisp/subr.el
      (read-passwd): Use read-string." (088be6fbd2).

    - Should it be mentioned in the manual?  delete-minibuffer-contents
      is documented of course, but not under "(elisp) Reading a
      Password".

    - a NEWS entry, maybe?  AFAICT C-u was never announced there.

(On a side-note, which might warrant another bug report, I wonder if
anything could be done to advertise read-passwd-map bindings more.
Sure, C-h b shows these bindings, but users have no reason to assume
that read-passwd has extra bindings in addition to regular minibuffer
ones, so why would they reach for C-h b?)

Let me know if this sounds like something useful.  If so, and if the
implementation is acceptable as-is, I'll work on documentation as soon
as I have a clearer idea of where I should add some (and whether some
documentation should be added for C-u as well).

Thank you for your time.

TODO:[[file:read-passwd-reveal.patch][attach]].
*** TODO Make cursor appearance context-dependent
E.g. add a customizable list of (PREDICATE . PROPERTIES) pairs; the
cursor gets the PROPERTIES from the first PREDICATE which matches.

Use-cases:
- org-speed-commands
- repeat-mode
*** completion
**** DONE [[bug:52169]] "Args out of range" when completing shell command
:PROPERTIES:
:Role:           author
:END:
*** font-lock
**** DONE [[bug:35476]] font-lock-{append,prepend}-text-property and anonymous faces
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
Started off as an [[orgmode:87r2a4ztt2.fsf@gmail.com][org-mode bug]]; Stefan applied a proper fix, then I
added in some refactoring and unit tests.
**** DONE [[bug:39597]] M-x occur adds fontification to fundamental-mode
:PROPERTIES:
:Role:           author
:END:
Fixed by Stefan & Juri.
*** fonts
**** DONE [[bug:41747]] Add default fontset setup for "Symbols and Pictographs Extended-A"
:PROPERTIES:
:Role:           author
:END:
Unfortunately, Symbola now comes with a [[https://dn-works.com/wp-content/uploads/2020/UFAS-Docs/License.pdf][non-free license]].

Cf. [[https://bugs.archlinux.org/task/58886][Arch Linux]], [[https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=897047][Debian]], [[https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/EC6I35ILFICKY5W5XTHYJC6UH36B2UQS/][Fedora]], [[https://www.mail-archive.com/gentoo-dev@lists.gentoo.org/msg88799.html][Gentoo]], [[https://github.com/NixOS/nixpkgs/pull/79679][NixOS]] and [[https://bugs.launchpad.net/ubuntu/+source/ttf-ancient-fonts/+bug/1824065][Ubuntu]] threads; also
[[https://en.wikipedia.org/wiki/Talk:Open-source_Unicode_typefaces#Symbola_font][Wikipedia]] thread.
**** DONE [[bug:51495]] Avoid fonts with incomplete coverage of MATHEMATICAL chars
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** imenu
**** TODO Add grouping/sorting predicates to ~imenu~ completion
So that ~imenu~ + ~icomplete-vertical-mode~ becomes a built-in,
buffer-less equivalent of ~imenu-list~.

Simple example of these predicates in action:

#+begin_src elisp
(completing-read
 "bleh? "
 (lambda (string pred action)
   (if (eq action 'metadata)
       (let ((sort-fn
              (lambda (candidates)
                (let* ((numbers (sort (mapcar 'string-to-number candidates) '<))
                       (evens (seq-filter 'cl-evenp numbers))
                       (odds (seq-difference numbers evens)))
                  (mapcar 'number-to-string
                          (if (and (car evens) (< (car evens) (car odds)))
                              (append evens odds)
                            (append odds evens)))))))
         `(metadata
           (cycle-sort-function . ,sort-fn)
           (display-sort-function . ,sort-fn)
           (group-function
            . (lambda (candidate transform)
                (if transform
                    candidate
                  (if (cl-evenp (string-to-number candidate))
                      "even"
                    "odd"))))))
     (complete-with-action action
                           (mapcar 'number-to-string
                                   (number-sequence 1 10))
                           string
                           pred))))
#+end_src

*** project
**** DONE [[bug:45765]] Change default-directory before prompting in project-compile
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** shr
**** DONE [[bug:39504]] Ensure faces of enclosing elements apply to <code> elements
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
Unfortunately debbugs.el does not handle multiple patches per message;
the first patch's diff was thus applied manually, with the second
patch's title (and a whole new message by the maintainer).
** Major modes
*** Custom
**** DONE [[bug:39074]] Horizontal line messes with variable value display in Custom buffers
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
**** DONE [[bug:51556]] Poor contrast of Customize SVG icons with dark backgrounds
:PROPERTIES:
:Role:           author
:END:
*** Dired
**** DONE [[bug:23284]] Cannot input 'Y' with dired-do-query-replace-regexp
:PROPERTIES:
:Role:           author
:END:
Fixed by Dmitry.
**** DONE [[bug:28969]] Confirmation prompt for wildcard not surrounded by whitespace
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
My commit message ran afoul of debbugs.el's =M-m= again.
*** ERC
**** TODO Support SASL authentication
Got bitten by this when trying to connect to irc.freenode.net from
=alyon-654-1-454-60.w109-213.abo.wanadoo.fr=:
#+begin_quote
ERROR from irc.freenode.net: Closing Link:
alyon-654-1-454-60.w109-213.abo.wanadoo.fr (SASL access only)
#+end_quote

This was requested in [[bug:29108]].  Lars closed this report because
while there is a third-party package to implement the feature, its
author did not respond when prompted for copyright assignment.

I did not have the patience to look at the package; instead I went
straight to [[https://ircv3.net/specs/extensions/sasl-3.1][the description of SASL on ircv3.net]] and the [[https://tools.ietf.org/html/rfc4616#section-2][RFC for the
=PLAIN= mechanism]], and cobbled this silly patch:

#+begin_src diff
diff --git a/lisp/erc/erc.el b/lisp/erc/erc.el
index 1d5506e281..0da677ac18 100644
--- a/lisp/erc/erc.el
+++ b/lisp/erc/erc.el
@@ -6052,7 +6052,17 @@ erc-login
                    erc-session-server
                    erc-session-user-full-name))
   (if erc-session-password
-      (erc-server-send (format "PASS %s" erc-session-password))
+      (progn
+        (erc-server-send "CAP REQ :sasl")
+        (erc-server-send "AUTHENTICATE PLAIN")
+        (erc-server-send
+         (concat "AUTHENTICATE " (base64-encode-string
+                                  (string-join (list
+                                                (user-login-name)
+                                                (erc-current-nick)
+                                                erc-session-password)
+                                               "\0"))))
+        (erc-server-send "CAP END"))
     (message "Logging in without password"))
   (erc-server-send (format "NICK %s" (erc-current-nick)))
   (erc-server-send
#+end_src

AFAICT this is enough to get me authenticated on Freenode.  It looks
like it's not too hard to implement?

Ideally Emacs's builtin =sasl.el= should be reused; that would let ERC
support better SASL mechanisms than =PLAIN=.
**** TODO Make timestamps robust vs window width changes
The =:align-to= spec for right timestamps inserts a fixed number of
spaces, so resizing windows messes up their display.  Eg. moving a
frame from one monitor to a smaller one, lines get wrapped
unnecessarily.

Don't know if the solution lies in (1) using one of the more
sophisticated forms of =:align-to= as shown in [[info:elisp#Pixel Specification][(elisp) Pixel
Specification]], (2) iterating over stamps on window-configuration
changes and updating their spec (3) just stop using right-aligned
stamps on the same line as messages.

(2) would not handle a buffer being shown in two windows with
different widths; (1) and (2) would not handle the window becoming too
narrow and requiring splitting the stamp from its message.  So leaning
toward (3).
**** TODO Play well with whitespace-mode
Last I checked, turning on whitespace-mode… hm.  That entry dates from
4 years ago.  Probably it broke ERC? 🤷
*** Gnus
**** DONE [[bug:40520]] Prevent duplicate thread titles
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
**** TODO [[bug:41530]] gnus-cloud-download-all-data does not mark articles as read
:PROPERTIES:
:Role:           author
:END:
**** TODO [[bug:42334]] gnus-thread-sort-functions vs. loose threads
:PROPERTIES:
:Role:           author
:END:
Closed by Lars because the way Gnus sorts things just isn't amenable
to fixing this right now.  Will probably need to take the plunge at
some point.
**** DONE [[bug:47026]] Allow Gnus summary faces to extend to end of line
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
**** DONE [[bug:49800]] Extend Gnus summary highlight faces by default
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** Ibuffer
**** TODO Elide with "…" if displayable
**** TODO Add human-readable size
With summary:
#+begin_src elisp
(define-ibuffer-column size-hr
  (:name "Size"
   :summarizer
   (lambda (strings)
     (file-size-human-readable
      (apply '+ (mapcar (lambda (s)
                          (get-text-property (1- (length s)) 'ibuffer-size s))
                        strings)))))
  (let ((sz (buffer-size buffer)))
    (propertize (file-size-human-readable sz) 'ibuffer-size sz)))
#+end_src

While at it, use this ~propertize~ trick to get rid of the "ewww"
comment in ~define-ibuffer-column size~.
*** Info
**** ~info-display-manual~
***** TODO redundant candidates
E.g. =foo= and =foo.info.gz=.
***** TODO invalid candidates
E.g. =gnutls-….png=
***** TODO use ~Info-additional-directory-list~
**** TODO redundant ~(if x (append x y) y)~
**** TODO redundant entries in ~Info-default-directory-list~ (Glenn's FIXME)
- 2000-12-14 "(Info-default-directory-list): If…" (e103ef954c)
  config-dir = "{configure-info-directory}/"
  config     = (list config-dir)
  config-dir ∈ standard-info-dirs
      ⇒ standard-info-dirs/config-dir ∨ config
- 2000-12-15 "(Info-default-directory-list): Don't…" (b6348438a2)
  config-dir ∈ standard-info-dirs
        ⇒ standard-info-dirs ∨ config
- 2012-05-26 "* lisp/paths.el (Info-default-…" (04188bb9ad)
  "why though"
*** Org
**** DONE [[orgmode:877efbgi4i.fsf@gmail.com]] Binding org-insert-todo-heading to M-S-RET
:PROPERTIES:
:Role:           author
:END:
Inspired by [[https://github.com/jrblevin/markdown-mode/pull/317][jrblevin/markdown-mode#317]]; fixed by Nicolas.
**** DONE [[orgmode:87r2a4ztt2.fsf@gmail.com]] Bug: Strike-through messes with heading face
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
Eventually reverted in favor of a better fix in [[bug:35476]].
**** DONE [[orgmode:877dxpazbo.fsf_-_@gmail.com]] Make RET and C-j obey ~electric-indent-mode~ in org-mode
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
Followup to an emacs-devel thread about reconciling org-mode
idiosyncrasies with Emacs core.
**** DONE [[orgmode:87blm8v09s.fsf@gmail.com]] Possible fix for :includes header argument in org-babel C source blocks
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
Stumbled on this regression before starting work on a completely
unrelated feature, and trying to get unit tests to pass.
**** DONE [[orgmode:87mu5mhm1u.fsf@gmail.com]] Failing tests
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
More unrelated tests that needed fixing before I could start working.
**** TODO [[orgmode:87mu62gvjk.fsf@gmail.com]] Setting org-todo-keywords through directory-local variables
:PROPERTIES:
:Role:           author
:END:
ACKed by Bastien; bump once 9.4 is released.
**** DONE [[bug:42184]] org-fontify-whole-heading-line does not work in emacs 27
:PROPERTIES:
:Role:           watcher
:PatchApplied:   t
:END:
An unfortunate fallout of the new =:extend= face attribute; fixed in
Org 9.4, which eventually landed in Emacs 27's maintenance branch
(cf. [[bug:43268]]).
**** TODO [[orgmode:87mu3ze52c.fsf@gmail.com]] Default description for abbreviated links
:PROPERTIES:
:Role:           author
:END:
Counter-proposal by Bastien.  Next step: ACK and get back to the
workbench.
**** TODO Skip checkbox width when filling list item
Currently items are filled like this:
#+begin_example
- [ ] lorem
  ipsum
#+end_example
Adding =\\[.\\]= to the regexp used in org-list-item-body-column
allows "ipsum" to be aligned below "lorem".
**** TODO [[bug:52587]] Prevent ~#+end_src~ background from extending to end-of-line when folded
**** TODO Add pandoc's ~auto_identifiers~ scheme for ID generation
~org-export-new-reference~ is the function which generates the current
ID, but it returns integers, so maybe the tweak should be made
somewhere above in the call stack?
**** TODO Make org-refile completion more similar to filename completion
I'd like to have org-refile completion work like =C-x p p=, =C-x p f=, and
Gnus's =B m=, where typing "bar" matches "x/bar".

Current org-refile config:
#+begin_src elisp
(setq org-refile-targets (list (cons nil (cons :maxlevel 10)))
      org-refile-use-outline-path t
      org-outline-path-complete-in-steps nil)
#+end_src

IIUC =C-x p= commands use ~project--file-completion-table~, which passes a
lambda to ~completing-read~ which returns ~(category . project-file)~
metadata, and…

#+begin_src elisp
(alist-get 'styles (alist-get 'project-file completion-category-defaults))
#+end_src

… says =substring=.  Likewise, ~gnus-summary-move-article~ uses
~gnus-completing-read~, which by default uses
~gnus-emacs-completing-read~, which sets ~completion-styles~ to
~gnus-completion-styles~, which includes =substring=.

Ergo: either I add =substring= to my ~completion-styles~, or I teach
~org-refile~ to use =substring=… by default? when ~path-complete-in-steps~
is nil? depending on a user option? 🤔
**** TODO Fix clock table not accounting for "descriptive links"
E.g. if a heading contains =[[http://long.url][Short description]]=,
the text that ends up in the clock table can be truncated to a shorter
length than the available width.

Assuming this is because there is a length-computing function that
does not account for ~org-link-descriptive~.
*** Programming modes
**** ~conf-mode~
***** TODO Tweak ~javaprop~ comments
As per [[https://docs.oracle.com/javase/10/docs/api/java/util/Properties.html#load(java.io.Reader)][the rationale]]:
#+begin_src diff
diff --git a/lisp/textmodes/conf-mode.el b/lisp/textmodes/conf-mode.el
index 57ec8a0428..91aabfb6d2 100644
--- a/lisp/textmodes/conf-mode.el
+++ b/lisp/textmodes/conf-mode.el
@@ -140,8 +140,7 @@ conf-unix-mode-syntax-table
 
 (defvar conf-javaprop-mode-syntax-table
   (let ((table (make-syntax-table conf-unix-mode-syntax-table)))
-    (modify-syntax-entry ?/  ". 124" table)
-    (modify-syntax-entry ?*  ". 23b" table)
+    (modify-syntax-entry ?! "<" table)
     table)
   "Syntax table in use in Java properties buffers.")
 
@@ -487,7 +486,7 @@ conf-javaprop-mode
   (conf-mode-initialize "#" 'conf-javaprop-font-lock-keywords)
   (setq-local conf-assignment-column conf-javaprop-assignment-column)
   (setq-local conf-assignment-regexp ".+?\\([ \t]*[=: \t][ \t]*\\|$\\)")
-  (setq-local comment-start-skip "\\(?:#+\\|/[/*]+\\)\\s *")
+  (setq-local comment-start-skip "^[ \t]*\\(?:#+\\|!+\\)\\s *")
   (setq-local imenu-generic-expression
 	      '(("Parameters" "^[ \t]*\\(.+?\\)[=: \t]" 1))))
 

#+end_src

**** ~python-mode~
***** DONE [[bug:51807]] Fix customization group of python-forward-sexp-function
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
***** DONE [[bug:52380]] ~run-python~ no longer focuses interpreter
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
**** ~sh-mode~
***** TODO Tweak indentation of continuation lines
Attempted to piggyback on Dario Gjorgjevski's [[bug:44592]] to fix these cases:
#+begin_src shell
${foo}bar \
    --arg
$(foo)bar \
    --arg

${foo}bar --arg1 \
          --arg2
$(foo)bar --arg1 \
          --arg2
#+end_src

Unfortunately my attempt was [[bug:50320][too naive]] and got reverted.  I'll have to
[[https://lists.gnu.org/archive/html/help-gnu-emacs/2021-09/msg00004.html][dig into SMIE]] more seriously…
** Minor modes
*** electric-pair
**** DONE [[bug:39680]] electric-pair-mode broken by undo
:PROPERTIES:
:Role:           author
:TestAdded:      t
:END:
Fixed by Stefan.  It took some time and effort, but I eventually
managed to write a unit test.
*** icomplete
**** DONE [[bug:38024]] icomplete sometimes fails to show completions after backward-killing words
:PROPERTIES:
:Role:           author
:END:
Fixed by João.
**** TODO [[bug:40152]] icomplete vs recursive prompts
:PROPERTIES:
:Role:           author
:END:
*** visual-line
**** TODO [[bug:31666]] Bad interaction between visual-line-mode and wrap-prefix on long lines
:PROPERTIES:
:Role:           watcher
:END:
Aka "visual-line-mode adds newlines before words that end up being
split anyway"; the problem occurs without wrap-prefixes.
** Themes
*** Modus
**** DONE [[https://gitlab.com/protesilaos/modus-themes/-/merge_requests/50][protesilaos/modus-themes!50]] Tone down ~icomplete-selected-match~
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
** ERT
*** TODO Noninteractive pitfalls
- [ ] testing faces vs redisplay (org-mode dir-local TODO keywords)
- [ ] post-command hooks (electric-pair + undo bug)
- [ ] current buffer (run-python)
* ELPA
** ada-mode
*** DONE [[bug:52167]] Preserve default value of ~project-read-file-name-function~
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
ada-mode pulls in uniquify-files, which unconditionally changes the
default value of ~project-read-file-name-function~.

- In uniquify-files: stop doing that; advertise this setting in the
  package commentary.
- In ada-mode: let-bind this variable in ~ada-find-file~.

Patches applied, though the maintainer did not preserve authorship nor
rationales from the commit messages.  For my own sanity:

- [[https://debbugs.gnu.org/cgi/bugreport.cgi?bug=52167;msg=5;att=1;filename=0001-Preserve-default-value-of-project-read-file-name-fun.patch][uniquify-files patch]] applied [[https://git.savannah.gnu.org/cgit/emacs/elpa.git/commit/?h=externals/uniquify-files&id=1d76b4f0e283afaff2be053d85f8726ffc3abd6e][here]];
- [[https://debbugs.gnu.org/cgi/bugreport.cgi?bug=52167;msg=5;att=2;filename=0001-Explicitly-bind-project-read-file-name-function.patch][ada-mode patch 1]]:
  - (ada-mode.el) applied [[https://git.savannah.nongnu.org/cgit/ada-mode.git/commit/?h=org.emacs.ada-mode&id=cd3bdbf6993d5f070795399c8f5ea538291c1d73][here]] (with further comment tweak), squashed
    with [[https://debbugs.gnu.org/cgi/bugreport.cgi?bug=52167;msg=5;att=4;filename=0003-ada-mode.el-ada-find-file-Support-future-Emacs-versi.patch][patch 3]];
  - (NEWS) applied [[https://git.savannah.nongnu.org/cgit/ada-mode.git/commit/?h=org.emacs.ada-mode&id=e5bc6c3c][there]];
- [[https://debbugs.gnu.org/cgi/bugreport.cgi?bug=52167;msg=5;att=3;filename=0002-ada-mode.el-ada-mode-menu-Prefer-ada-find-file.patch][ada-mode patch 2]]: not applied;
- [[https://debbugs.gnu.org/cgi/bugreport.cgi?bug=52167;msg=5;att=4;filename=0003-ada-mode.el-ada-find-file-Support-future-Emacs-versi.patch][ada-mode patch 3]]: applied [[https://git.savannah.nongnu.org/cgit/ada-mode.git/commit/?h=org.emacs.ada-mode&id=cd3bdbf6993d5f070795399c8f5ea538291c1d73][here]], squashed with [[https://debbugs.gnu.org/cgi/bugreport.cgi?bug=52167;msg=5;att=2;filename=0001-Explicitly-bind-project-read-file-name-function.patch][patch 1]].
** adaptive-wrap
*** DONE [[bug:41810]] Fontify wrap-prefix
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** TODO More fontifying?
1. Start ~markdown-mode~.
2. Write some ~> long line~ of quoted text.
3. Start ~visual-line-mode~ and ~adaptive-wrap-prefix-mode~

~>~ markers are not fontified.
*** TODO Take ~display~ property into account
1. Start ~markdown-mode~.
2. ~markdown-toggle-markup-hiding~.
3. Write some ~> long line~ of quoted text.
4. Start ~visual-line-mode~ and ~adaptive-wrap-prefix-mode~

The ~wrap-prefix~ does not match the opening ~>~'s ~display~ property.
** debbugs
*** DONE [[bug:38551]] bind to RET rather than return
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** TODO autoload debbugs-gnu-emacs-release-blocking-reports
I prefer this over the Org variant, which is autoloaded.
*** TODO Make ~debbugs-gnu-apply-patch~ smarter
To avoid accidents like [[bug:28969]], [[bug:39504]], [[bug:41810]], and
[[bug:57733]].
** diff-hl
*** DONE [[https://github.com/dgutov/diff-hl/issues/142][#142]] Weird interaction between diff-hl-flydiff-mode and org-indent-mode
Eventually reported back to Emacs core in [[bug:41584]] and fixed by Eli.
** which-key
*** DONE [[https://github.com/justbur/emacs-which-key/issues/314][#314]] Some keymap names no longer show up
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
* MELPA
** magit
*** DONE Helping ~magit-ediff-dwim~ read my mind
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
[[https://emacs.stackexchange.com/q/18200/10209][Reported on StackExchange]] first, then [[https://github.com/magit/magit/commit/f65b1bddd31070e6012fe91a7d845f748c40b901][applied by tarsius]].
*** DONE [[https://github.com/magit/magit/pull/2928][magit/magit#2928]] magit-split-range: fix handling of "..@{upstream}"
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** DONE [[https://github.com/magit/magit/pull/3720][magit/magit#3720]] Make magit-blame-echo actually echo the commit summary
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** DONE [[https://github.com/magit/magit/pull/4043][magit/magit#4043]] Fix git-commit fontification when comments contain brackets
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** DONE +[[https://github.com/magit/magit/pull/4207][magit/magit#4207]]+ [[https://github.com/magit/magit/pull/4213][magit/magit#4213]] Autoload magit-file-mode-map correctly
:PROPERTIES:
:Role:           author
:END:
The initial state of affairs:

- ~global-magit-file-mode~ says ~:init-value t~, but that has no
  effect.
- We need to move ~magit-file-mode~ (and ~magit-blob-mode~ while we're
  at it) to a new, dedicated library, and either
  - let users customize ~global-magit-file-mode~ to t, which will
    DTRT,
  - autoload the form that enables the mode if the variable is set.

It's already possible to do either, but it slows down startup
considerably.  Hopefully moving the mode to a file that does not
~(require 'magit)~ will mitigate this?

[[https://lists.gnu.org/archive/html/help-gnu-emacs/2020-09/msg00130.html][For context.]]  Eventually closed in favour of [[https://github.com/magit/magit/pull/4237][tarsius's own solution]].
** markdown-mode
*** DONE [[https://github.com/jrblevin/markdown-mode/pull/124][jrblevin/markdown-mode#124]] Prevent spurious bold fontification
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
*** DONE [[https://github.com/jrblevin/markdown-mode/issues/172][jrblevin/markdown-mode#172]] Spurious bold/italic in inline code
:PROPERTIES:
:Role:           author
:TestAdded:      t
:END:
*** DONE [[https://github.com/jrblevin/markdown-mode/issues/223][jrblevin/markdown-mode#223]] Bold markers mess up italics detection
:PROPERTIES:
:Role:           author
:END:
*** DONE [[https://github.com/jrblevin/markdown-mode/pull/252][jrblevin/markdown-mode#252]] Fix markdown-inline-code-face's :inherit attribute
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** DONE [[https://github.com/jrblevin/markdown-mode/pull/317][jrblevin/markdown-mode#317]] Bind markdown-insert-list-item to M-RET rather than M-<return>
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:TestAdded:      t
:END:
** page-break-lines
*** DONE [[https://github.com/purcell/page-break-lines/pull/20][purcell/page-break-lines#20]] Fix disabling when major mode is derived from one of page-break-lines-modes
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
** paradox
*** DONE [[https://github.com/Malabarba/paradox/pull/125][Malabarba/paradox#125]] Invert :inherit order for paradox-mode-line-face
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
*** DONE [[https://github.com/Malabarba/paradox/pull/160][Malabarba/paradox#160]] Fix "Invalid face reference" warnings for custom-button-mouse
:PROPERTIES:
:Role:           author
:PatchApplied:   t
:END:
* COMMENT File-local variables
#+LINK: bug     https://debbugs.gnu.org/
#+LINK: orgmode https://orgmode.org/list/%s/t/#u
# Local variables:
# org-property-format: "%-16s %s"
# end: